Source encoder for video pictures having a quantizer whose stepsize is controlled in an incremental manner
Patent 5028999 Issued on July 2, 1991. Estimated Expiration Date: 
July 2, 2008. Estimated Expiration Date is calculated based on simple USPTO term provisions. It does not account for terminal disclaimers, term adjustments, failure to pay maintenance fees, or other factors which might affect the term of a patent.
July 2, 2008. Estimated Expiration Date is calculated based on simple USPTO term provisions. It does not account for terminal disclaimers, term adjustments, failure to pay maintenance fees, or other factors which might affect the term of a patent.Patent ReferencesSystem comprising a preliminary processing device controlled in accordance with an amount of information stored in a buffer Adaptive picture image compression system DPCM system with rate-of-fill control of buffer occupancy Adaptive graylevel image compression system Patent #: 4725885 InventorApplicationNo. 171726 filed on 03/22/1988US Classes:348/419.1Coding element controlled by buffer fullnessExaminersPrimary: Groody, James J.Assistant: Kostak, Victor R. Attorney, Agent or FirmInternational ClassH04N 007/12Foreign Application Priority Data1987-03-27 DEDescriptionBACKGROUND OF THE INVENTION1. Field of the Invention The invention relates to a source encoder for video pictures having an encoding unit, and including a control unit which supervises the filling level of a buffer store connected in cascade to an encoding unit and which in dependence upon that filling level, controls a quantizer included in the encoding unit. 2. Description of Related Art A source encoder with these features is disclosed in an article by G. Lutz et al. (Lutz, G., Speidel, J. und Streicher, E.: Der Videokonferenz-Codec 2Mbit/s-Ein System zur Uebertragung von Bewegtbildern und Grafiken, Sprache und Daten mit 2Mbit/s. PKI Technische Mitteilungen (1985), Vol. 1, pages 50-60). FIG. 1 shows a basic circuit diagram of the prior art source encoder. In this FIG. D denotes a data source producing data from video pictures for an encoding unit C, in which a code conversion of the video data is effected in a bit-saving line code. The encoding unit C is preceded by a pre-processing unit PP. The pre-processing unit PP arranges the video data in blocks: i.e. the data of picture elements which are located in square picture sections. The encoding unit C includes inter alia a picture store in which every T seconds all the blocks associated with a video picture (video pictures must here be understood to mean fields or frames) are stored via a switch S and are intermediately stored until the moment at which they are processed. In addition it also includes a controllable quantizer by means of which the number of bits is controlled by coarse or fine quantization, by which the information components of the blocks are encoded with a more or less great loss of information. This encoded information is written at a bit rate which fluctuates in time in a buffer store PS and, read therefrom at a constant bit rate and transferred to a receiver. The filling level F of the buffer store PS is continuously supervised by a control unit K. The encoding unit C, the quantizer and the switch S are controlled such, in dependence on the filling level F, that the buffer store PS can neither overflow nor be emptied completely. If, for example, the filling level F of the buffer store PS increases, then the control unit K provides that the quantization stepsize is continuously increased so that consequently the quantization gets coarser. If the filling level decreases the quantization is refined. If in spite of the coarsest quantization the filling level F increases to above the maximum value Fmax, the switch S is opened for at least an integral multiple of T seconds, i.e. at least one video picture is omitted from the video picture sequence at the transmitter end (frame or field sub-sampling). This fact is transmitted to the receiver by means of a appropriate information, the video picture received last being repeated on the display screen of this receiver, instead of the missing video picture. On reaching the filling level Fmax the overall encoding circuit C is inoperative for T seconds; then the filling level F of the buffer store PS decreases as it is emptied at a constant bit rate until a lower level Fo. Then the encoding unit C resumes operation, and more specifically--as the filling level increases again--with ever increasing quantization stepsizes, so increasingly coarser quantization, until the filling level Fmax is reached again or all the blocks stored in the picture store of the encoding unit C have been processed. When the filling level Fmax has been reached again, the encoding unit is stopped again for T seconds. When the stored blocks have been processed, the switch S is closed as soon as the subsequent integral multiple of T seconds after its opening has ended. The quality of the video picture displayed on the display screen at the receiver end in the prior art encoder needs improvement on the following grounds: During the encoding of a video picture stored in the picture store of the encoding unit C, all the values for the quantization interval are in certain circumstances even cycled-through many times. Consequently, fluctuations between the highest and the poorest quality of individual picture areas occur within a video picture. At the receiver end these circumstances result in that the displayed picture gives an unacceptable total impression for the observer for a short time. SUMMARY OF THE INVENTION Consequently, the invention has for its object to provide an encoder of the type defined in the opening paragraph with which an improved picture quality can be obtained at the receiver end compared with the known encoder. This object is accomplished in that the control unit controls the quantizer in dependence upon the period of time after the maximum filling level of the buffer store has been reached and/or in dependence on the actual value of the quantization stepsize, in such a manner that during the entire operation the value of the quantization stepsize only passes through adjacent values. Advantageous embodiments of the invention are described hereinafter. The invention and how it can be put into effect will now be described with reference to the accompanying Figures. BRIEF DESCRIPTION OF THE DRAWING FIG. 1 shows a basic circuit diagram of a source encoder according to the prior art mentioned in the preamble and FIG. 2 shows a time diagram for the filling level of a buffer store and the value of the quantization stepsize of a quantizer. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 has extensively been described in the introductory part of this specification. For the sake of completeness it is mentioned here that FIG. 1 does not indicate the clock supply, not even a suggestion thereof, as it is not an object of the invention. For a detailed structure of the source encoder a person skilled in the art is referred to, for example, the prior art mentioned in the preamble and the literature cited there. FIG. 2 shows a possible variation of the filling level F of a buffer store and also the value QI of the quantization stepsize of a quantizer used in the encoding unit C of FIG. 1. Time instants are plotted on the time axis of FIG. 2 with intervals of T seconds. At each of these instants the preprocessing unit PP of FIG. 1--the switch S is consequently closed in normal operation--conveys in normal operation the data of a frame to the picture store of the encoding unit C. Consequently, 1/T is the picture sequence frequency shown in Hertz; in the present example this frequency has a value of 10 Hz for pictures having 288 lines and 360 picture elements per line. As can also been seen from FIG. 2, the filling level F (solid-line curve) of the buffer store PS twice reaches its maximum value Fmax, more specifically at the instants t1 and t2. At these instants the total activity of the encoder C of FIG. 1 is stopped for T seconds (100 ms) (processing stop); more specifically, no data are then entered into the buffer store PS of FIG. 1, so that this store is emptied in the present example at a bit rate of 64 kBit/s. Consequently after 100 ms the content of the buffer store PS has been reduced by 6400 bits; then the encoder C is made operative again and the filling level F increases again from the value Fo to the value Fmax. The encoder C is now stopped again and the buffer store is only emptied during 100 ms. If finally the picture stored in the picture store located in the encoder C has been processed in this manner, then the switch S, which was opened at the instant t1, is closed again and a new picture is transferred to the picture store. The pictures which were provided by the preprocessing unit PP at the instants 3T, 4T, 5T and 6T, are not transferred to picture store due to the fact that the switch S is open. Also in the case the switch S is in the open state, the value QI of the quantization stepsize of the quantizer is changed during the processing of the picture stored in the picture store of the encoder by the control unit, provided the filling level of the buffer stores changes. At a high filling level the quantization becomes coarser, consequently the quantization stepsize get higher. At a maximum filling level Fmax the value QI is highest, namely QI max. Whilst in the prior art the value QI only depends upon the filling level of the buffer store PS, that is to say that after a processing stop of the encoder C the quantizer operates with much finer quantization. In accordance with the invention the value QI is now dependent upon other parameters to such an extent that the value of the quantization stepsize QI is not subjected to sudden and great changes. It is namely these changes during the encoding operation of a picture which result in an unacceptable picture quality. In the present example the value QI only depends on the filling level F of the buffer store and on the period of time elapsed after the maximum filling level of the buffer store. To simplify the explanation, an auxilary variable v is determined by the following equation: ##EQU1## wherein tn is the instant at which the filling level F has reached the value Fmax for the last time after t seconds have elapsed and c symbolizes a predeterminable, positive time constant. Typical values of c are T/10 seconds, i.e. 10 ms in the present example. If now the value of the quantization stepsize is controlled such in dependence on the auxiliary variable and the filling level F, that QI(v,F)=(1-v)g(Fmax) v g(F), (1) wherein g(F) is a "steadily" and monotonously increasing function of F, then the value QI of the quantization stepsize cannot "jump"; it is a "steady" function of the two variables v and t, respectively, and F. The quotation marks are necessary since all the quantities are indigital representation with a finite number of bits and consequently the notions "steady" and "jump" cannot be used in their original meaning. A possible variation of QI(v,F) is plotted in FIG. 2 (dot-and-dash line), therein QImax =g(Fmax) and QIo =g(Fo). The variation shows the "steadiness" of QI. It should be noted that during the operation stop of the encoder C, the value QI need not to be defined, as it is not needed. It is only of importance that during the operation of the encoder C the value QI does not "jump", but only passes through adjacent values, independent of the fact whether the operation is interrupted by a pause or not. |
